ML19290A134
| ML19290A134 | |
| Person / Time | |
|---|---|
| Site: | Crane  |
| Issue date: | 08/09/1979 |
| From: | Jerome Murphy NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| To: | NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| Shared Package | |
| ML19290A132 | List: |
| References | |
| NUDOCS 7910170355 | |
| Download: ML19290A134 (5) | |
Text
{{#Wiki_filter:. 1 NN O t y, pg UNITED STATES (, p, NUCLEAR REGULATORY COMMISSION 5-ef WASHINGTON, D. C. 20555 E o .....,o August 9, 1979 MEMORANDUM FOR FILES FROM: Joseph A. Murphy Probabilistic Analysis Staff Office of Nuclear Regulatory Research
SUBJECT:
LOSS OF FEEDWATER TRANSIENTS IN B&W REACTORS A brief scoping analysis of the likelihood of core melt accioents initiated by loss of feedwater transients for a typical Bcbcock & Wilcox designed reactor has been prepared. It should be noted that this analysis is based on a typical B&W design as it existed before the modifications and improved operator training required by the bulletins and orders issued since the occurrence of the TMI-2 accident. The analyses performed are limited in scope and are based on point estimates of failure rates. Error propagation techniques have not been employed and the uncertainty of the final results have not been determined in a formal manner. To the extent possible, this analysis has relied on previous analyses performed under the Methodology Applications Program sponsored by the Probabilistic Analysis Staff and the Reactor Safety Study (WASH-1400). The information from the Methodology Applications Program is preliminary and has not yet been published. To analyze loss of feedwater transients at Babcock & Wilcox designed reactors, an event tree has been constructed (Figure 1, attached). For ease of understanding, symbology of the Reactor Safety Study is used to designate failure states as noted at the top of the event tree (i.e., K represents reactor trip fails, etc.). Each sequence in this event tree is discussed below: a. Sequence 1 (TM) - This is the expected loss of feedwater transient. Last year it had an average frequency of approximately 3 per reacter year at B&W reactors, as noted in NUREG-0560. The number of feedwater transients at an individual reactor varies from zero up to an extreme of 10 or so per year, depending on how long the plant has been in commercial operation (highest number in first year or two). The loss of feedwater causes a reduction in the heat remelai capability of the secondary system which, in turn, causes a pressure rise in the primary system. This pressure r,ise causes the pressurizer relief valve to open and resultant conditions cause the reactor to trip. The auxiliary feedwater system is initiated upon the loss of main feedwater and has the capability of removing decay heat. Thermal balance is restored by the auxiliary feedwater action, relief valve 7910170365 2223 291
. operation and reactor trip. The pressure decreases, the relief valve recloses, and the plant is maintained in a hot shutdown condition. A typical FSAR for a Babcock & Wilcox design (TMI-2) using a conservative analysis predicts that a peak reactor coolant system pressure (2500 psig) above the relief valve setting (2250 psig) will be reached. Thus, the relief valve (and possibly a safety valve) will be called upon to open in most feedwater transients. I C b. Sequences 2, 3, and 4 (TMQ1, TMQ1 C, TMQ1 ) - These sequences follow the same path as sequence 1 (TM) except that the relief valve, after opening, fails to reclose. This failure of the valve to reclose is essentially a small loss of coolant accident located in the vapor space of the pressurizer. The consequences of the transient depend on the success of ECCS in coping with this small LOCA. If ECCS functions as in sequence 2 (TMQ ), no core damage should result. If ECCS 1 is unable to deliver its design capability (either because of hardware failures or human intervention), degraded core condii;icgs involving a significant degree of fuel damage could result (sequence 3, TMQ1C ). In the limit, if :CCS fails to function, core melting can occur (sequence 4, TMQ1 ). The likelihood C of a feedwater transient coupled with failure of the relief valve to reciose can be estimated from available data to be approximately 6x10-2 per reactor year for an average B&W plant. (Main feedwater is lost approximately three times per reactor year and B&W plant operating experience indicates the lhelihood of the relief valve failing to reclose is approximately 2x10-2 Wr demand. ) The unavailability of the emergency core cooling system in this sequence is difficult to determine. Unavailabilities for the ECCS operating in the high pressure injection mode f r the liqyid space LOCAs have been e 'timated to be in the range of lx10- to 4x10-J per demand at other plants. However, the experience at B&W plants with vapor space loss of coolant incidents indicate that there is a relatively large likelihood that the operator might take actions which would degrade or disable the ability of the ECCS to perform its function in such a vapor space LOCA. The probabi-lities of ECCS failure, P, and of ECCS degraded performance, PC, have not 4x10 q/ demand.stimatedbutitisblearthat(PC+PCl) is significantly greater tha I been 3 Considering the experience with stuck open relief valves in B&W plants, the combined conditional probability of PC + P 1 may exceed 0.1 C based on subjecti matedtobe6x10gejudgment. Thus the likelihood of sequence 2 ( is esti-(1-P -P I); the probability of sequence 3 (TMQ1 {MQj)he TMI-2 C C C ), t which leac'; to core melt is estimated to be 6x10 jP.s estimated to be 6x10-'P j ; a
- sequence, c
C Note that the summat on of the prt.babilities of sequences 2, 3, and 4 shoulb be approximately 6x10-{/RY. Sequence 5 (TMPj) - This sequence is similar to sequence 1 (TM) except the c. relief val e does not open. However, a safety valve does open to relieve pressure and properly recloses. No fuel damage is predicted or this sequence. Assuming an unavailability of the relief valve to open of 1 sed on WASH-1400, it ha a probability of occurrence of approximately 3x10 ' I d. Sequences 6, 7, and 8 (TMP Q, TMP Q C, TMPjQ C) - These sequences are similar tosequences2,3,and4,biscusseh2underb.above. In these sequences,_ 2 however, the relief valve fails to open, a safety valve opens and then fails ,j,' - 2223 292
to reclose producing a vapor space LOCA similar to what occurred at TMI-2. As in sequences 2, 3, and 4, the degree of core damage is dependent on the degree of ECCS operability. In these sequences, however, since the safety valve lines do not have block valves which can terminate the LOCA, the ECCS must function in both the injection and recirculation modes. For a liquid space LOCA, previous analyses would indicate an unavailability range for the combined injection and recirculation modes of 6x10-3 to lx10-2/RY. However, as in b,above, experience indicates that the human may intervene in such a manner as to impede syr'.em performance. Using the values indicated in b. and a failure rate for a safety valve to reclose of 10-2/ demand based on operating experience as assessed in WASH-1400, the following probabilities are estimated: Sequence 6 (TMPjQ2 = 3x10-4(1-PCI-P ) - no core damage C Sequence 7 (TMPjQ CI = 3x10-4 PCI) - degraded core 2 Sequence 8 (TMP Q C = 3x10-4 P ) - core melt j2 C Thesummationofthg/RY. probabilities of sequences 6, 7, and 8 should not exceed approximately 3x10-Sequence 9 (TMP h) coolant system will overpressurize to some extent.In this e. open. Theprim$r While high pressures may be experienced, it is unlikely that the pressure would be sufficient to cause primary system failure. Thus, no significant fuel damage would be expected. Based on the assumptions stated above and a failure rate of the safety valves to open of apprgximately 10-5/ demand, this sequence has a calculated likelihood of 3x10 /RY. WASH-1400 indicated the unavailability of a safety valve on demand was approximately 3x10-5, Toaccountforpossibgecommonmodefailuresbetweenthetwosafety valves a value of 10- was used here. f. Sequence 10 (TML) - In this sequence there is a prolonged loss of all feedwater. The primary system pressure increases to the safety valve setpoint and decay heat is dissipated by boiloff from the primary system. Some makeup to the primary system can be obtained to c#fset the boiloff using the high pressure injection system; however, because of the pressure increase, this flow will generally be inadequate because the HPIS pumps will be near their shutoff head and ccre melt will result. Under some circumstances, however, it may be possible to prevent core meltdown. The simplified event tree presented in Figure 1 does not show all possible states which might exist if all feedwater were lost. Specifically, there is some chance that if a relief valve were to stick open, it would pass sufficient steam to depressurize the primary coolant system to some extent. This would permit an increase in ECCS flowrate which might prevent complete core melt b'Q. - i 2223 293
e from occurring. The unavailability of the auxiliary feedwater system has been analyzed in the Reactor Safety Study and in the Methodology Applications Program, unavailability range froe.10'ginary results indicate po previously referenced. Preli to 6x10-3/ demand.. Using this range, t e likelihood 2x10-2, depending on the actual unavailability of the auxiliary feedwater system. There is some likelihood, not considered herein, that main feedwater could be recovered in sufficient time to prevent core damage. This recovery time is on the order of 10-15 minutes because steam may not be available to power the turbine driven main feedwater pumps after the steam generators dry out. Since the likelihood of recovery is small in B&W plants, it has not been considered. g. Sequence 11 (TMK) - This sequence is a severe anticipated transient without scram (ATWS). It will produce primary system pressures in excess of 4000 psi which could lead to failure of the reactor coolant system. NUREG-0460 estimated the unavailability of a Babcock & Wilcox-designed reactor protection system as 3x10-b/ demand. Based on an initiating event frequency of three per year and this value for failure to scram, the likelihood of core melting as a result of this sequence is assessed to be approximately 9x10-5/RY. Sequences 10 and 11 could be expanded to include the effects of relief and safety valve failures. Since all sequences so identified will lead to core melt (with the exception noted in the discussion of sequence 10), they have been eliminated here for ease of communication. Eummary tombining the results obtained above, the likelihood of experiencing cgre degradation or undergoing core melting appears to be dominated by sequence 3(TMQjC8) and 4(TMQjC). is clear their combined likelihoo? of These have not been directly quantified but ij/ reactor year and could exceed ence is significantly higher than 2x10-occurg/RY. Sequence 10, TML, could be dominant depending on the reliability 6x10-of the auxiliary feedwater system. It is worth noting that sequence 10 (and the approximate probabilities noted) applies to all PWRs. 410seph A. Murphy Probabilistic Analysis Sta" Office of Nuclear Regulatory Research
Attachment:
Loss of Feedwater Event Tree - Babcock & Wilcox Design
- .. M -
2223 294 s,..
Loss of Feedwater Event Tree - Babcock & Wilcox Design TM K L Pi P2 01 02 C Loss of Reactor Auxiliary Relief Safety Relief Sa fety (C -degraded Seq. Core ECCS l Comments Feedwater Trip feedwater Valve Open valve Open Valve Closes Valve Closes performance) Condition LOCA 1 OK j l 7,' f_ 2 OK / (see s ?.x 10,~2/ D, r { e ,Pcl text 3 Degraded Y Vapor Space i < r a a a LOCA (see v Pr text) 4 Melt j a' ! l l' Yes S OK N N^ 6 OK N ] (see (r3 ~10~2/D Prl text) 7 Degraded j 5, Vapor Space s10-2/D I g N (see LOCA Pr text) 8 Melt <10 5/D LJ1 9 No Helt? -Yb& High Primary System Pressure 6x10-3-10-5/D (see. :t) 10 Helt (sea text) s3x10-5/D s, 11 Melt Very High Primary No Note: TMI-2 Accident Sequence Cross Hatched System Pressure figure 1 .}}